Study population, compliance, adverse events and secondary outcomes
For this study, 18 ileostomy subjects were screened of which 16 (Additional file 1, Supplementary Table 2,additional file 4,
Supplementary CONSORT flow diagram)were enrolled. Fifteen subjects completed the protocol (see Materials and Methods).In general compliance was high and no serious adverse events were reported during this trial (Additional file 1, Supplementary Table 1, Supplementary Figure 2).
The primary aim was to determine the impact of the consumption of fermented dairy products on the small intestine microbiota composition and activity. In addition, intestinal permeability and SCFA levels in ileostoma effluent samples were determined at the start and end of each of the intervention periods. Notably, neither permeability nor SCFA amount or composition was significantly affected by any of the interventions (Additional file 1, Supplementary Figures 3, 4 and 5).
Longitudinal metataxonomic analysis of the small intestinal microbiota
Longitudinal microbial composition in ileostomy effluent was analysed by metataxonomic analysis, which was successful in 390 (>90%) of the 432 collected effluent samples (27 per individual). The samples obtained from two subjects had the lowest success rate, which for at least one of the subjects appeared to be due to the low DNA recovery in the samples (Additional file 1, Supplementary Table 3). There was no difference in the successrate in obtaining data across the different periods of the trial (Additional file 1, Supplementary Table 4). Initial principal component analysis (PCA)analysis of the microbiota data obtained (Additional file 1, Supplementary Figure 6) revealed thatall samples obtained from one subject strongly deviated from the rest. This subject was the only individual with a Kock’s pouch rather than the standard ileostoma, which led us to exclude this subject in further analyses as a biological outlier. In the remaining samples, 9270 operational taxonomic units (OTUs) were identified, representing 258 genera, and 9 phyla.
The majority of the sequences were assigned to Firmicutes (78.6%) and Proteobacteria (11.9%), followed by Actinobacteria (6.3%) and Bacteroidetes (3.1%). The average microbiota composition per subjectrevealed a high degree of variation among the subjects (Supplementary Figure 6). Redundancy analysis revealed that almost half (46.6%) of the overall microbiota composition at species level was explained by inter-subject variation (Additional file 1, Supplementary Figure 6). In addition, longitudinal composition analysis per subject revealed remarkable time-dependent fluctuation during the study period, although composition appeared more stable in some individuals (Additional file 1, Supplementary Figure 6). Nevertheless, intra- and inter-subject comparison of Bray-Curtis dissimilarity during the study showed a higher degree of variance between subjects than within subject (Additional file 1, Supplementary Figure 7).These metataxonomic analyses expand insights gained in previous studies[14, 15], by confirming the high degree of variance of the SImicrobiota composition between and within volunteers, while establishing that a personal microbiota signature is recognizable despite temporal fluctuation.
Impact of fermented dairy product consumption on the SI microbiota composition
The fermented milk products consumed during the study contained either Lacticaseibacillus rhamnosusCNCM I-3690 or the yogurtstarter culture bacteria Streptococcus thermophilus CNCM I-1630 and Lactobacillus delbrueckii subsp. bulgaricus CNCM I-1519. The placebo control product was unfermented, acidified milk. In agreement with the relatively low numbers of L. bulgaricus(see Materials and Methods)in the yogurt product, this species remained undetected in the effluentmicrobiota following consumption. Thereby, detection of product derived bacteria (PDB) in the effluent microbiota was restricted to the abundant bacteria inthe respective products, i.e., L. rhamnosus, and S. thermophilus. Importantly, thesebacteria were not detected in any sample obtained outside of the product-specific intervention periods, showing PDB transiently inhabit the small intestinal nicheand rapidly disappearonce consumption is stopped and establishing that the two-week washout suffices to avoid microbial carry-over between intervention periods, which is also confirmed by the absence of a detectable difference between the last samples obtained during the wash-out periodand those obtained during the run-in (Additional file 1, Supplementary Figures 1,8).
In samples obtained during the intervention periods, PDB corresponding to the intervention product could be detected (Figure 1). On average, L. rhamnosus and S. thermophilus constituted 8.5, and 5.2% of the total microbiota in the effluent samples collected during the 2 intervention weeks. However, highly variable relative abundances were observed when comparing intra- and inter-personal samples. Illustratively, 3 subjects had a PDB average relative abundance below 1% (max 5%) and the PDB could not be detected (<5.7E-03) in some of the samples, whereas 3 other subjects had a PDB average abundance above 10%, including the striking maximum relative abundance of L. rhamnosusat 88.2% in one of the samples. These observations exemplify the strong oscillation of relative abundance of the PDB in all subjects (Figure 1), supporting the dynamic nature of the small intestinal microbial community.
Despite the large impact of the PDB in some samples, the alpha diversity of the effluent microbiota remained unchanged during the intervention periods as compared to the run-in, although alpha diversity varied substantially between the subjects (Additional file 1, Supplementary Figure 9). The impact of the intervention on the microbiota composition was analysed by subject corrected redundancy analysis (RDA), revealing a significant effect of the intervention at both genus and species level, albeit that these effects only explained a small fraction of the overall variation (Figure 1). The drivers of the significant separation of the intervention-specific samples were the PDB (Figure 1). Moreover, species-level beta diversity analysis revealed that the samples obtained during the interventions were significantly different (Additional file 1, Supplementary Figure 10), which was supported by the inclusion of the PDB in the coremicrobiome of the effluent samples during the respective intervention periods. Notably, the PDB addition to the core microbiome was the sole difference detected in this analysis (Additional file 1, Supplementary Figure 11).
Interventions significantlyaffect the SI endogenous microbiome
The predominant effect of the product interventions on the small intestine microbiota composition appears to be the presence of PDB in samples collected during the intervention period. To investigate the possible effect of the interventions on the endogenous microbial community, the PDB-related OTUs were removed followed by recalculation of the relative abundances of the remaining OTUs. The resulting dataset was used for subject corrected RDA analysis demonstrating that samples could still be significantly separated on basis of the intervention period in which they were taken (Figure 1), albeit that samples displayed a large overlapped, and only a very low amount of variation could be explained (0.44% and 0.48% at species and genus level, respectively). The 11 endogenous species associated with the different interventions were detected by empirical analysis of digital gene expression (EdgeR)differential abundance analysis, correcting for subject ID (Additional file 1, Supplementary Table 5). However, none of these species were prevalent among the majority of subjects or present at relative high abundance (max prevalence: 27%, max abundance 1.5%). These findings indicate that the intervention-impacton the endogenous microbiota was poorly conserved and barely detectable, which agrees with the minimal amount of explained variation found by RDA.
Endogenous Peptostreptococcaceae abundance is correlated with the relative abundance of PDB during intervention
The analyses above indicated that the average relative abundance of the PDB during the intervention periods differed substantially per subject(Figure 1). Remarkably, high congruency was observed between the subject-specific average relative abundance of S. thermophilus and L. rhamnosus during the respective interventions, suggesting PDB colonization efficiency is very subject specific but independent of the PDB species.This led us to investigate whether the endogenous baseline microbiota could explain these differences in abundance of L. rhamnosus and S. thermophilus. The corresponding RDA analysis revealed that while several bacterial families appeared to be enriched in subjects that displayed high relative abundance values for the PDB, only the Peptostreptococcaceae were negatively associated with the detected abundance of the PDB (Figure 2). The significant negative correlation between PDB colonization efficiency and endogenous Peptostreptococcaceae relative abundance could be confirmed by univariate analysis. Notably, despite substantial variation within and between subjects, we could not detect a correlation between the DNA yield obtained from ileostoma effluent samples and Peptostreptococcaceae or PDB average relative abundance (Additional file 1, Supplementary Figure 12), indicating that the abundances of neither Peptostreptococcaceaenor PDB is associated with a difference in bacterial density. Intriguingly, high Peptostreptococcaceae relative abundance was correlated with a lower microbiota diversity (Figure 2).
Colonization efficiency of PDBcorrelates with elevated expression of carbon fermentation pathways in the small intestine microbiota
The microbial activity in ileostomy effluent was analyzed by metatranscriptome analysis in the samples obtained on the first and last days of the intervention periods.Functional metatranscriptome mapping (FMM) of the small intestine microbiome were obtained for each sample by genome,protein and pathway mapping.
Inter subjectFMM differences were the predominant source of variation, explaining almost 25% of the overall FMM variance (Additional file 1, Supplementary Figure 13).This is supported by the higher Bray-Curtis distance between FMM data obtained for different subjects, compared to longitudinal FMM data from a single individual irrespective of the intervention period (Additional file 1, Supplementary Figure 13).Nevertheless, subject-corrected RDAanalysis revealed that the interventionsdid significantly affect the FMM and could explain 4.64% of the overall variance in the FMM data (Additional file 1, Supplementary Figure 14). The differentially expressed microbial pathways that underly the intervention-associated effect on the FMM were identified using EdgeR differential expression analysis, with subject ID as a co-variate. This revealed that only expression of the L-rhamnose degradation I pathway was significantlyincreased in the FMM associated with the L. rhamnosusintervention relative to the placebo intervention (Additional file 1, Supplementary Table 6), while no FMM effects were found when comparing yogurtand placebo interventions. Notably, comparative FMM analysis of samples obtained during the yogurt and L. rhamnosusinterventions revealed the increased expression of 8 carbon- and fermentation- associated pathways during the L. rhamnosus intervention, including the L-rhamnose degradation I pathway(Additional file 1, Supplementary Table 6).These results show that the consumption of the L. rhamnosusfermented product associates with elevated expression of carbon metabolism pathways in the small intestinal microbiota, which is especially apparent for the L-rhamnose degradation pathway.
Endogenous microbiome pathway activity patterns influencethe colonization efficiencyofPDB
To investigate whether variations in the FMM data were associated with the highly variable and subject-specific relative abundance of PDBduring the intervention periods, RDA analysis was performed using the FMM determined during the intervention periods to explain the average relative abundance of PDBduring their respective interventions(Figure 3). A strikingly strong enrichment was found for various bacterial pathways related to amino acid metabolism in samples displaying the lower PDB relative abundance, which was contrasted by an association of glycolytic pathway expression (i.e., Glycolysis IV in EcoCyc) with the higher PDB relative abundance (Figure 3). These findings were confirmed using the multivariate association with linear models (MaAsLin2) analysis, which also pointed to positive associationsof PDB abundance and glycolytic pathway activity (Glycolysis IV and Glycolysis III), while expanding this to the Stachyose degradation pathway (Figure 3). MaAsLin2also confirmed the negative association between PDB abundance and the expression of the majority (9 out of 12) of amino acid biosynthesis pathways identified in FMM (Figure 3).Subsequently, metagenomic phylogenetic analysis (MetaPhlAn) of the FMM data enabled the phylogenetic classification of these differentially expressed pathways, revealing that the activated amino acid biosynthesis pathways were assigned to a broad range of bacterial families in a scattered manner (Figure 3), encompassing most of the microbial families encountered within the small intestinal ecosystem (30 out of 41). In this context, it is striking that the family of the Peptostreptococcaceae is one of the few bacterial taxa that negatively correlated with the amino acid biosynthesis pathways in the FMM.Taken together, our results highlight how the endogenous microbiota composition and its predominantly active energy metabolism could explain the individual-specific PDB colonization efficiency, specifically identifying the Peptostreptococcaceaefamily as a microbial group that prominently reflects, and is likely involved in,these individualized effects.